1 s2.0 016541019390034D Main

Journal of Accounting and Economics 16 (1993) 407-445.
North-Holland
The investment opportunity set and

accounting procedure choice
Preliminary
evidence*
Received November 1990. final version received July 1992
This paper provides evidence on the cross-sectional relation between firms investment opportunities, their debt and compensation contracts. their size and financial leverage. and their accounting
procedure choices. This evidence is important. because previous studies hypcthesize !!?ut the link
between firms investment oppor!-unities and their accounting choices helps explain extant results on
the size. debt/equity. and bonus plan hypotheses. However. while 1 find that firms rnvestment
opportunities do affect the nature of their coIltracts, I also find that the traditionalexplanations for
accounting choice are important after controlling for the effects of the investment opportunity set.
1. Introduction
This paper provides evidence on the cross-sectional relation between firms

investment opportunities (hereafter, the investment opportunity set or just the
ios), the nature of their debt and compensation contracts, their size and
financial leverage, and their accounting procedure choices. The principal
Cnt~~spc~r&~~ to: Douglas J. Skinner. School of Business Administration. University of Michigan,

Ann Arbor, MI 48109-1234, USA.
*I thank Jeff Abarbanell. Don Anderson, Joy Begley, Vie Bernard, Harry DeAngelo. Linda
DeAngelo. Tony Greig. Pat OBrien. and Jerry Zimmerman for helpful comments on earlier
versions, and Li Li Eng, James Myers, and Karren Skinner for help with data collection. Special
thanks are due to Tom Lys (the referee) and Ross Watts (the editor) for comments and suggestions
that helped me to improve the paper substantially. Financial support from the KPMG Peat
Marwick Foundation and the University of Michigan Business School is also gratefully acknowledged. All errors and omissions are my responsibility.
The term nature of contracts refers to the extent to which contracts(i) restrict the actions of the
contracting parties and (ii) use accounting numbers to measure performance and/or specify actions
and,or prescribe restrictions, Y elaborate further on this in section 2.
01654101/93~$06.00
( 1993
Elscvier Science Publishers B.V. All rights reserved
408
D.J. Skinner, Investment opportunities and accounting choice
empirical conclusions are as follows:

(1) Consistent with extant research, I find that larger firms are more likely to
select income-decreasing accounting procedures, that more highly-levered firms
are more likely to select income-increasing accounting procedures, and that
firms with accounting-based bonus plans are more likely to select incomeincreasing accounting procedures, than are other firms. Importantly, I also find
that these relations remain largely unaffected when I include measures of the ios
in the regressions - the traditional explanations are robust to controlling for
the effect of the ios.
(2) I document that the ios affects the nature of firms debt and compensation
contracts. Specifically, firms with relatively more assets-in-place are more likely
to employ (i) accounting-based debt covenants in their public debt contracts and
(ii) bonus plans that tie the bonus directly to accounting earnings. Thus managers of firms with more assets-in-place have larger incentives, given the nature
of the contracts in place in these firms, to select income-increasing accounting
procedures. This is evidence of an indirect relation between the ios, through its
effect on contracting, and accounting choice.
(3) There is limited evidence of a direct association between the ios and
accounting choice. Specifically, in some accounting choice regressions the ios
proxy variables have explanatory power with respect to accounting choice over
and above the traditional variables that proxy for managerial opportunism.
Although this evidence suggests that the ios affects the accepted set of accounting procedures, it should be interpreted cautiously given the correlation that
exists between the ios and the nature of firm contracts.
This evidence contributes to the literature on positive accounting theory in
several ways. First, although there has been speculation in the literature that the
ios affects accounting procedure choice [Watts and Zimmerman (1986, 199Q)],
there is little evidence on the relation between the ios and accounting choice.
The only direct evidence of a relation between the ios and accounting choice (of
which I am aware) is in Zimmer (1986), who examines whether Australian real
estate developers capitalize interest costs. The present study is therefore the first
to provide evidence about the ios/accounting choice relation for a large, randomly-chosen cross-section of firms.
Second, most previous studies view accounting choice as a function of
managers* incentives to behave opportunistically, given contracts in place. [See
Holthausen and Leftwich (1983) or Watts and Zimmerman (1986, ch. 11) for
reviews.] These studies typically regress measures of accounting choice on firms
debt-to-equity ratios (to proxy for the incentives that debt contracts provide to
managers) or on the existence of a bonus plan (to proxy for managers incentives
under those agreements), and interpret the relations they observe as providing
evidence on the extent to which managerial opportunism drives accounting
choice. For example, a positive relation between debt-to-equity ratios and
D. J. Skinner,
Invesrmenr
opportuniries
and accounting
choice
409
the use of income-increasing
accounting procedures is usually interpreted as

evidence in favor of the hypothesis that managers choose income-increasing
accounting procedures to loosen debt-cr-*nvenant constraints.
If firms investment opportunities systematically affect the types of contracts
that are written, the inferences that these studies draw cocbld be confounded by
an omitted variables problem [Watts and Zimmerman (1990)].* To address this,
I include measures of the ios in accounting choice regressions along with the
traditional measures such as firm size, financial leverage, and a bonus plan
indicator variable. In addition, I collect data on the nature of sample firms debt
and compensation contracts, and report on how the ios affects the nature of
these contracts across firms. Thus, this paper examines whether the relation
between the ios, firm contracts, and accounting choice helps explain results
already in the literature. Alternative explanations are important, for as Leftwich
(1990, p. 41) emphasizes, even if the existing tests provide unambiguous evidence of associations in the data, there is little agreement on interpretations of
those correlations.
Finally, this evidence increases our understanding of observed accounting
practice. For example, firms in the same industry tend to adopt similar accounting practices. Because it is likely to be the case that the nature of investment
opportunities vary across industries more than they do within industries, the
ios-accounting choice linkage provides a simple and direct explanation for this
phenomenon. More generally, the ios-accounting choice linkage helps explain
why firms accounting procedure choices appear to depend on factors such as
their production and investment, financing, and operating decisions [see, e.g.,
Foster (1986, ch. 5)].
To analyze cross-sectional accounting choice data for a large number of firms,
I have to make simplifying assumptions. For example, I characterize alternative
accounting procedure choices at a given point in time as either incomeincreasing or income-decreasing - in practice the earnings effects of particular
accounting choices tend to reverse over time. (Although with inflation this type
of reversal will only occur for firms that are shrinking, and so should not
systematically characterize the firms in my sample.) In addition, managers are
likely to consider the effect of alternative accounting methods on more than just
the level of earnings (e.g., under income smoothing hypotheses, managers
attempt to reduce the variability of earnings). While more specific research
settings avoid such ambiguities [e.g., see Zimmer (1986) J, the results from these
studies are more difficult to generalize to a broad cross-section of firms.
Thus, while there are limitations of the evidence that I provide, this evidence
2Along similar lines, Ball and Foster (1982) and Holthausen and Leftwich (1983) had originally
criticized these types of accounting choice studies because of the likelihc: ,d that the size, debt/equity,
and bonus plan proxy variables that they employed are likely to be correlated with many other
things, including industry membership.
410
nonetheless increases our understanding of extant accounting practice, particularly when combined with the evidence from studies that examine smaller
samples of firms in particular economic circumstances (such as Zimmers).
The next section of the paper summarizes existing theory on the relation
between the ios and accounting choice, and emphasizes that it is not obvious
a priori how the ios affects the accepted set of accounting procedures. In
addition, I provide a rationale for why we expect a relation between the ios and
the nature of firm contracts, and so why there is also likely to be an indirect link
between the ios and accounting choice. Section 3 provides details of the sample.
Section 4 discusses the proxy variables that I use to measure the ios, and
documents how the ios covaries with firm size, financial leverage, and long-term
financial performance. Section 5 provides evidence of how the ios affects the
nature of firms bonus plans and debt contracts. Section 6 then provides
evidence of variation in firms accounting choices, and documents how this
variation is related to variation in these firms size, financial leverage, debt and
compensation contracts, and investment opportunities. Section 7 provides a discussion and summary.
2. The ios-accounting choice linkage

This section describes two links between the ios and accounting choice.
section 2.1 I discuss why we expect to observe a direct relation between the
and the accepted set of accounting procedures. Section 2.2 discusses how the
is likely to affect the nature of firm contracts and thus managers incentives
select particular accounting procedures after contracts are in place, and
provides an indirect link between the ios and observed accounting choice.
2.1.
In
ios
ios
to
so
The ios and the accepied set qf accourzting procedures
Watts and Zimmerman (1986, 1990j isolate two principal forces that determine firms accounting procedure choices. First, certain accounting procedures
(the accepted set) evolve through time and emerge as best practice - these are
the accounting procedures that cost-effectively resolve the firms agency problems ex ante. This perspective on accounting choice is also known as the
efficient contracting perspective [e.g., Holthausen ( 1990)].3 Second, because it
3An early example of this argument is Watts (1977, p. 61) who draws a linkage between the agency
costs of debt, the nature of firms assets, and the requirement that managers set aside profits for
renewal, repairs, maintenance or depreciation of existing assets. Leftwich (1983) argues that the
accounting choices that have evolved as best practice in private lending agreements (as exemplified
by their existence as boiler-plate type provisions) are those procedures that minimize the costs of
the agency relationship between the firm and its private lenders. Zimmer (1986) provides an example
in the context of Australian real estate developers of how accounting procedure choices can result
from contracting solutions that are rx ~tte efficient.
D.J. Skinner,
Inwstment
opportunities
and accounring
choice
411
is costly to restrict managerial choice entirely, managers choose particular

accounting procedures from among the accepted set ex post, i.e., after the
contracts are in place. Managers who do this to make themselves better off at
the expense of other stakeholders are said to act opportunistically. The extant
literature focuses primarily on managers incentives to make opportunistic
choices.4
In delineating an accepted set of accounting procedures, those appointed to
monitor the firms managers face a tradeoff. On one hand, managers accounting
choices must be restricted to some extent - otherwise contracts denominated in
accounting nuqsbers are ineffective in restricting manager behavior. For
example, compensation contracts would do little to motivate managerial performance if earnings-based management compensation agreements did not
specify the accounting choices on which the earnings calculation was based.
On the other hand, however, if there are efficiency-based costs and benefits to
particular accounting choices, managers are likely to best know which accounting procedure choices maximize the value of the firm. In other words, managers
are likely to have specic knowledge about which accounting choices are optimal
from the point of view of all claimants. For example, managers are likely to have
the best information about which accounting procedures minimize the firms
potential costs in the political/regulatory
process, or about which accounting
methods provide the best way of motivating employees (thus ex post choices are
often efficient as well). Therefore, it is not efficient to contractually restrict
managerial choice entirely, given this as well as the costs of monitoring and
enforcing restrictions on accounting choice.5
The relative costs and benefits of restricting managerial choice are likely to be
different for different firms. In particular, these costs and benefits - and so the
eficient ser of accounting procedures - are likely to vary across firms as a function of their investment opportunities. For example, Watts and Zimmerman
(1986, pp. 360-361) argue that the assets of growth firms are more difficult to
observe because they are primarily represented by future investments. As a result, contracts based on these less readily observed values provide managers
with greater flexibility to behave opportunistically ex post, SOthat growth firms
accepted sets of accounting procedures would likely restrict managers ability to
choose income-increasing accounting procedures ex ante. (On the other hand,
4Choices that are made after contracts are in place are not always the result of managerial
opportunism - managers also make choices that are KYpost efficient from the point of view of the
firms claimants.
In addition to Watts and Zimmerman this general point has previously been made by Demski,
Patell, and Wolfson (1984) who argue that the delegation to managers of the choice from among the
set of acceptable (accounting) alternatives can best be understood as efficient, equilibrium behavior
(p. 17) and that by both allowing mangers freedom of choice with respect to accounting methods
and compensating t/tern according/y, owners can capitalize upon managers local expertise (p. 31.
emphasis in original).
412
the accounting choices that are optimal for growth firms probably depend on
as-yet unobservable outcomes, so it may be more costly to restrict accounting
choices in the current period for these firms.)
While it is clear that the ios will affect both the costs and benefits of writing,
monitoring, and enforcing contracts, it is difficult to arrive at more specific
predictions about how exactly the ios will affect the accepted set of accounting
procedures. As Watts and Zimmerman (1986, p. 359) note: Development of
a complex model of the accepted procedure set is difficult since it would involve
modeling the demand for contracting as a monitoring device and the cost of
alternative methods of monitoring. As a result, I adopt the more modest
objective of providing descriptive evidence on the relation between the ios and
observed accounting choicle.
2.2. The indirect relation between the ios and accounting choice
In addition to its direct eXects on the accepted set, the ios is likely to affect the
types of contracts that the firm writes with various claimants. To illustrate this
in a concrete way, I next analyze how the ios affects the nature of management
compensation hnd debt contracts, respectively.
2.2. I. Management compensation agreements
Management compensation agreements help reduce the conflict of interest
between corporate managers and stockholders; these plans are designed to
motivate managers to maximize firm value [Smith and Watts (1982)]. Because
the costs and benefits of monitoring and motivating managers depend on the
nature of the firms investment opportunities, the structure of management
compensation agreements will vary across firms as a function of the ios. Smith
and Watts (1991) argue that the actions of managers of firms with relatively
more assets-in-place are less costly to monitor than the actions of managers of
firms whose value is comprised largely of growth opportunities. In addition,
managers of firms with relatively more growth opportunities are likely to be
allowed more decision-making discretion because these managers have better
information about the firms investment opportunities than the firms stockholders; i.e., managers of growth firms ari: likely to have relatively more specific
knowledge than managers of firms with relatively more assets-in-place. Consequently, Smith and Watts predict that growth firms are more likely to use
incentive compensation schemes that tie management compensation to
measures of firm performance (such as accounting earnings or stock price).
In addition, Smith and Watts (1991) argue that accounting numbers are
poorer performance measures for firms with relatively more growth opportunities because of conservatism in accounting: the need for objective and verifiable
numbers limits the extent to which accountants are willing to recognize income
that depends on uncertain future events. [Consistent with this, Collins et al.
6). J. Skinner. Investment oppomnities and accounting choice
413
(1992) provide evidence that stock prices lead accounting earnings to a greater
extent for firms with relatively large amounts of new and intangible assets.] This
countervailing measurement effect implies that incentive compensation schemes
that rely on accounting performance measures are less likely to be used in firms
with growth opportunities, offsetting the arguments above. Stock-based incentive plans are likely to be more prevalent in growth firms, since these plans
achieve the incentive-alignment benefits without incurring the accounting
measurement costs.6
This discussion is summarized in fig. 1. The figure illustrates the two countervailing effects of the ios on the nature of managerial compensation - firms
with more assets-in-place are: (1) less likely to use incentive compensation
(because their managers optimally have less decision-making discretion) and (2)
more likely to use accounting earnings numbers in compensation contracts
(because these numbers are relatively better performance measures in these
firms).
Smith and Watts (1991) and Gaver and Gaver (1993) provide evidenc+cthat is
consistent with this. Smith and Watts present industry-level evidence that firms
with growth opportunities are (i) less likely to use accounting-based bonus plans
than other firms and (ii) more likely to use stock-option plans. Gaver and Gaver
(1993) replicate the Smith and Watts analysis for 237 growth firms and 237
nongrowth firms, and find that their growth firms are more likely to use
stock-option plans, but no more likely to use earnings-based bonus plans, than
their nongrowth firms. The fact that the evidence indicates that growth firm: are
more likely to use stock-option plans, but is inconclusive with respect to which
firms are more likely to use bonus plans, is consistent with the existence of these
two opposing effects.
Overall, it is difficult to make a clear a priori prediction about the relation
between the ios and the use of accounting-based bonus plans because of the two
countervailing effects, However, the evidence in Smith and Watts (1991) and the
evidence below (table 3) suggests that, at least as an empirical matter, firms with
relatively more assets-in-place are more likely to use bonus plans that rely on
accounting numbers. Consequently, in fig. 1 I indicate that firms with more
assets-in-place are more likely to employ earnings-based bonus plans. If this is in
Sibson (1990, pp. 342-343) prescribes for fast-growing enterprises: There is continuous change
in these firms. Therefore, compensation systems must be simple and highly adaptable to
change . . . In fast-growing companies, you dont have annual bonus plans. Performance standards
would be changing frequently. Also, some plans, like management bonus plans, can be divisive in
a fast-growth environment.
7See also Clinch (1991) who finds that high-R&D firms are more likely to use stock-option plans
than low-R&D firms, but that there is no apparent difference in these firms propensity to use
earnings-based bonus plans, and Lambert and Larcker (1987). who find that firms with relatively
high rates of sales and asset growth place more weight on market returns and less weight on
accounting numbers in determining managerial compensation than do other firms.
414
D.J. Skinner,
Investment
opportunities
and accounting
choice
Less likely to
use incentive
compensation
More hkely to use

earnings-based
bonus plans*
Firm with more

assets-in-place
More likely to use

income-increasing
accounting
procedures
__t
L
Accounting numbers
are relatively good
measuresof
performance
I
Opportunism Linkage
Fig. 1. The link between the investment opportunity set, incentive compensation, accounting
earnings-based bonus plans, and accounting procedure choices. (*The fact that firms with more
assets-in-place are more likely to use earnings-based bonus plans is an empirical statement based on
the evidence in this and o:her studies.)
fact the case in general, and if the contractual terms of bonus plans provide
managers with incentives to make income-increasing
accounting procedure
choices, then I expect that firms with more assets-in-place are more likely to use
accounting-based bonus plans, so that their managers are more likely to select
income-increasing accounting procedures ex post. This is shown as the last
linkage in fig. 1.
Debt convers-ts
are written to reduce the conflict of interest between the
firms stockholders and bondholders [Smith and Warner (1979)]. Smith and
Warner describe four principal sources of conflict between bondholders and
stockholders, two of which have implications for the ios-accounting
choice
linkage.
*Unlike the accruals choices that Healy (1985) documents, managers accounting procedure
choices cannot be altered every period depending on where the firms accounting income falls
relative to the upper and lower bounds in bonus plan contracts; it is relatively costly to change
accounting procedures [see e.g., Sweeney (199l)J Therefore, the relation between the existence of
bonus plans and managers* accounting procedure choices is likely to depend on whether incomeincreasing procedure choices will, on average over a period of years, increase the managers bonus.
Based on the relatively strong evidence in favor of the bonus plan hypothesis, this is probably the
case. However, this is ultimately an empirical question and in the tests that follow I provide evidence
about the relation between the ios and the terms of bonus plans for a subsample of firms with
available data.
The other two sources of stockholder-bondholder
conflict are dividend payment (under which
the stockholders opportunistically reduce the value of the bondholders claims by reducing the firms
level of investment to increase the dividend) and claim dilution (under which stockholders
opportunistically dilute the value ofexisting bondholdersclaims by issuing more debt of the same or
higher priority). Neither of these conflicts relates directly to the nature of the firms assets, and so
neither of these conflicts is linked to the firms ios.
D. J. Skinner, hwesrment opportmifies and accounting choice
415
1. The Underinuestment Problem. As discussed above, while firm managers

have little discretion with respect to the value of assets-in-place, they can readily
affect the value of growth opportunities. With risky debt outstanding, Myers
(1977) shows that situations can arise where managers (acting on shareholders
behalf) decide not to undertake these positive net present value investments
because most of the payoffs go to bondholders. To mitigate this underinvestment problem, firms generally only issue risky debt that can be supported by
assets-in-place. The analysis thus predicts that, other things equal, the larger the
fraction of firm value represented by assets-in-place, the higher the firms
financial leverage.
2. The Asset-Substitution Problem. Asset substitution occurs when stockholders opportunistically substitute higher variance assets for lower variance assets,
once the debt has been issued. Asset substitution will transfer wealth to the
stockholders if the debt has been issued and priced on the assumption of
low-variance assets. The extent to which asset substitution creates agency costs
will vary across firms as a function of their ios. If a firms ios is comprised largely
of assets-in-place, it is relatively easy to prevent asset substitution - the assets
are fixed assets which are relatively easy for outsiders to monitor (external
auditors, for example, can easily verify the continued existence and upkeep of
fixed assets such as land, buildings, plant, and equipment). On the other hand, if
firm value is largely comprised of intangible growth opportunities, asset substitution is likely to be more difficult to prevent. Therefore, as with underinvestment, firms with relatively more growth opportunities are less likely to issue
risky debt (other things held constant).
D-+1,
Aerinvestment
and asset substitution suggest that firms with relatively
ULll I,
UIUI&
___A
more assets-in-place will be more highly levered than firms whose va!ue is
comprised principally of growth opportunities. Gaver and Gaver (1993), Long
and Malitz (1985), and Smith and Watts (1991) all provide evidence that is
consistent with this prediction. This relation is shown as the upper part of the
first linkage in fig. 2.
If firms whose value is made up principally of growth opportunities are less
likely to issue debt, these firms are also less likely to have accounting-based debt
covenants. This follows because: (i) these firms have less debt so that bondholders do not require the same degree of protection, (ii) the cost of these restrictions, in terms of inhibiting these firms future investment decisions, is likely to
be larger for these firms, and (iii) as argued above, accounting numbers are likely
to provide relatively poor information for monitoring stockholders of these
firms. These relations are shown in fig. 2.
Press and Weintrop (1990) provide evidence that firms that use accounting-based debt covenants are more highly levered than firms that do not have these constraints. If firms with more
assets-in-place are also more highly levered, this evidence supports the view that the ios affects the
likelihood that firms will have accounting-based debt covenants.
416
Higher financial
leverage
More likely to use

accoumic, ja.xd
debt covenanls
Firm with more

assets-in-place
Accounting numbers
are relatively good
measuresof
performance
More likely IO use

income-increasing
accounting
procedures
*I
OpportunismLinkage
Fig. 2. The link between the investment opportunity set, financial leverage, accounting-based debt
covenants, and accounting procedure choices.
Extant evidence on the debt/equity hypothesis indicates that larger

debt/equity ratios are associated with the use of income-increasing accounting
procedures, and suggests that managers choose income-increasing accounting
procedures to loosen debt-covenants constraints. However, if the ios affects both
financial leverage and the extent to which firms utilize accounting-based debt
covenants, these studies potentially suffer from a correlated omitted variables
problem.
It is difficult to discriminate between these explanations empirically: is it
opportunism that explains the observed relation between accounting choice and
leverage, or is it that the ios determines both leverage and the firms accounting
choices? Press and Weintrop (1990) provide evidence that, even after controlling for managers incentives under accounting-based debt covenants, leverage is
still significant in explaining cross-sectional variation in accounting choice. As
they note, one expianation for this result is that leverage is proxying for the ios in
these regressions. By including ios measures directly in these regressions, the
tests in this paper provide evidence on whether or not this is true, and, more
generally, on the extent to which the ios determines accounting choice conditional on the incentives that managers have under contracts in place.
2.2.3. Summary
I have argued that the ios affects the accounting procedures used by firms
indirectly. This occurs because the ios affects the nature of the contracts written
In section 2.2.1, the opportunism to which I was referring involved wealth transfers from the
firm to its managers under managerial compensation contracts. In this section, I am referring to
opportunism under debt agreements that involves wealth transfers from bondholders to stockholders that are implemented by managers (who are, following Smith and Warner, assumed to act in the
best ir,:eres& of the stockholders). In both cases managers choose income-increasing accounting
procedures .I effect the wealth trpn$ers.
D. J. Skinner,
hcestnzennr opporruniries
und accounting
choice
417
between the firm and its stakeholders, and because - other things equal - the
nature of these contracts, in turn, affects the accounting procedures that managers select ex post. The ceteris paribus assumption here is important, because
the effects of the ios on the accepted set of accounting procedures may offset
managers opportunistic incentives (this is an empirical question). For example,
if firms with relatively more assets-in-place are more likely to use earnings-based
bonus plans, so that managers of these firms have larger incentives to make
income-increasing accounting choices cx post, it may also be that these firms
accepted sets of accounting procedures are more likely to restrict managers
ability to make these choices ex ante.
3. Sample selection
To obtain
a broad cross-section
of firms I select all firms on the 1989

Compustat II PST jik (Compustat) that have the requisite data available. To
focus primarily on unregulated industrial firms, I exclude firms in the following
four-digit SIC categories: (i) 100-1400 (mining firms), (ii) 2911 (petroleum firms),
(iii) 4011 (railroads), (iv) 4911-4991 (utilities), (v) 6021-8744 (financial, medical,
and service firms). Many of these firms do not report depreciation or inventory
accounting choices (because they do not have material depreciable assets and/or
inventories), which are my primary interest here.12
Table 1 lists the number of firms selected in each two-digit SIC code industry,
along with the number of firms available for selection in each industry. The
data-collection constraint is the availability of these firms accounting procedure
choices (financial and stock market data being relatively cheaply available on
Compustut and CRSP). I obtain these data from firms annual reports, primarily
from the first footnote where firms generally disclose their summary of significant accounting policies. The firms in the sample are those that are incorporated in the United States, have 1987 annual reports available in the University of
Library, and disclose their accounting
Michigan Kresge D**-;-cc
Duallavuu
_4dministration
procedure choices. l3 The sample comprises 78% of the relevant population of
Compustut firms (504 of the 650 available firms). The coverage varies by industry
from a minimum of 60.8% (steel firms) to a maximum of 100% (several industries).
Although mining and petroleum firms have both fixed asset: and inventories, they tend to
account for these assets in unique ways. and so are excluded from !he sa:nple. For example. because
mining firms often recognize revenue at the point of production, their inventories die usually
recorded at amounts that exceed cost, i.e., at market value. And both mining and petroleum firms
typically use some type of depletion base to allocate the cost of their fixed assets to expense, making
it difficult to compare their depreciation methods directly to those of other industrial firms
(moreover, for many of these firms the important choice is between full cost and successful efforts
accounting, not between straight-line and accelerated depreciation).
Consistent with Can~pustat, I define a 1987 annual report as one for the year ended over the
6/30/87 to 5131188 period.
418
D.J. Skinner,
Inoestmenr
opportunities
and accounting
choice
Table 1
Summary of SIC-defined industry representation of 504 sample firms with available data in 1987.
Industry description
SIC codes
Number of
sample firmsa
Heavy Construction
Food Products
Textile Mills
Finished Apparel
Millwork, etc.
Furniture
Paper Products
Newspaper Publishing
Periodicals, Books
Chemicals
Drugs, Pharmaceuticals
Plastics
Footwear
Cement
Blast Furnaces (Steel)
Nonferrous Metals
Metal Tools, Parts, etc.
Machinery
Computing Equipment
Electrical Machinery
Semiconductors
Motor Vehicles
Aircraft/Aerospace
Instruments & Controls
Air Transport
Radio & TV Broadcasting
Retail - Dept. Stores
Retail - Grocery
Retail - Apparel
Retail - Fast Food
1600- 1799
2000-2070
2200-222 1
2300-2330
2421-2451
2510-2522
2600-263 1
2711
2721-2731
2800-2829
2834
3080- 3089
3140
3241-3270
3312
3330-3334
341 f-3460
3510-3537
3570-3571
3600-362 I
3674
3711
3721-3760
3812-3861
4512
4833
531 l-5331
5411
5651
5812
10
34
13
18
11
12
22
11
6
24
27
15
10
11
14
11
33
20
19
14
14
7
28
51
10
6
20
9
6
18
12
47
17
25
14
15
23
13
6
33
39
18
10
II
23
14
40
?2
23
19
15
10
32
75
15
7
26
13
6
27
504
650
Total firms
Population
of firmsb
The number of firms in a given industry group in the final sample. This represents those firms in
the population (see footnote b below) that are incorporated in the United States, have 1987 annual
reports on tile at the Universi!y of Michigan Kresge Business Administration Library, and that
disclose their accounting choices.
bThe populatio n represents all firms in a given industry group on the 1989 annual Compustot
industria! tape.
4. Size, leverage, and investment opportunity set proxy variables

4.1. The meusures
The accounting procedure choices that I seek to explain are those reported at
fiscal year-end 1987. Therefore, the financial variables are measured for each
firm as the average of the year-end values for 1985 through 1987. The data are
D. J. Skinner, Incesmtent opporruniries and accounting choice
taken from annual

separately.
Compustat. 1 discuss measures
419
of each of the constructs
4.1. i. investment opportunity*set measures
The construct that I seek to measure here is Myers (1977) notion of assetsin-place vs. growth opportunities. I use three variables to capture this construct.
In addition to these three variables, I include a measure of the riskiness of these
firms assets.
First, to measure assets-in-place (i.e., those assets whose ultimate value does
not principally depend on future discretionary investment by managers), I calculate the ratio of the book value of gross property, plant, and equipment (PP&E)
to the value of the firm. The intuition behind this measure is that past investments in property, plant, and equipment can be characterized as assets-in-place.
This measure is increasing in the proportion of assets-in-place to value.
Second, as an additional proxy for growth opportunities vs. assets-in-place,
I use the firms research and development (R&D) expense deflated by net sales.
The rationale here is that investments in R&D yield expected payoffs that form
part of managers private information and that the value of these investments is
difficult for outsiders to measure reliably. Moreover, R&D expenditures comprise, at least to some extent, discretionary expenditures and so are like the
growth options that Myers describes [Dechow and Sloan (1991)]. Long and
Malitz (1985) and Smith and Watts (1991) also use R&D expenditures to
measure Myers notion of growth opportunities.
Finally, I also include Tobins 4 ratio, which is defined as the market value of
the firm divided by the replacement cost of its assets. This variable increases with
the proportion of firm value represented by intangible assets. Both Lindenberg
and Ross (1981) and Merck, Shleifer, and Vishny (1988) provide arguments
and/or evidence that Tobins 4 increases with the proportion of value represented by intangibles [in fact Merck, Shleifer, and Vishny (1988, p. 299) use
R&D and advertising variables to control for observable measures of intangible
assets that affect $1.
Following Lindenberg and Ross (1981), I estimate the replacement cost of
these firms assets as the sum of the replacement cost of property, plant, and
equipment (PP&E), the replacement cost of inventories (measured as the book
value of inventories plus the LIFO reserve), and the book value of other assets.
I estimate the replacement cost of PP&E using the procedure outlined by Hall et
al. (1988) for the NBER R&D master file data:
1. Estimate the average age of the PP&E by dividing 1987 year-end accumulated depreciation by the 1987 depreciation expense (this assumes straight-line
depreciation).
2. Multiply the 1987 net book value of PP&E by ihc ratio of the GNP
deflator for fixed nonresidential investment in 1987 to this same deflator AA
420
D.J. Skinner,
Inaesrtnent
opportunities
and accounting
choice
years before 1987, where AA is the average age of the assets from 1. [This
assumes that the replacement cost of PP&E grows, on average, at the same rate
as the GNP deflator for fixed nonresidential investment. Hall et al. (1988) and
Lindenberg and Ross (1981) use the GNP deflator for fixed nonresidential
investment in similar calculations.]
I then divide the market value of the firm (i.e., the size measure described
below) by this estimate of the replacement cost of the assets to yield an estimate
of q.
To proxy for the risk of these firms investment opportunities, I estimate their
asset betas. To calculate an asset beta, I first estimate the beta of each firms
stock, and then unlever this number by multiplying by the equity-to-value ratio
discussed below [see Hamada (1972)]. This estimation procedure assumes that
the firms debt is approximately risk-free.14
To summarize, I use four variables to characterize the ios for these firms: one
measure that is positively related to assets-in-place/firm value (gross PP&Eto-value), two measures that are positively related to growth opportunities/firm
value (R&D-to-sales and q), and a measure of risk (asset beta).
4.1.2. Firm size
and
financial leverage
I measure the size of the firm as the sum of the market value of equity and the
book value of debt (current liabilities plus long-term debt). Two measures of
financial leverage are employed, one based on book values and the other on
market values. The book value measure is the ratio of the book value of
long-term debt to the book value of total assets (debt-to-assets), while the
market value measure is the ratio of the market value of equity to firm size,
calculated as described above (equity-to-value). The debt-to-assets measure is
increasing and the equity-to-value measure is decreasing in financial leverage.
4.1.3. Accounting ROA
Firms past accounting performance is used since there is reason to believe
that accounting procedure choice is related to how well or badly firms are
performing: firms that are poor performers are more likely to select incomeincreasing accounting procedures [see, for example, Lilien, Mellman, and Pastena (1988) and Sweeney (1991)]. It is important to control for accounting
14The betas are estimated for most (305) firms using five years of monthly CRSP data through
December 1987. Some firms are not available on the CRSP monthly returns files, however, and so
for these firms (134) I use 500 days of daily return data from the CRSP daily returns file. I thus have
beta estimates for a total of 439 sample firms. I use montly beta estimates where available because
these are less susceptible to the measurement error that arises because of nonsynchronous
trading
than are daily beta estimates.
D.J. Skinner,
Inresmenl
opportunities
and amounting
choice
421
performance because, for example, firms recent accounting performance may

be correlated with their ios: firms for which assets-in-place comprise a large
fraction of value may have performed systematically better (or worse) than
other firms. The particular measure that I use here is an average (over the
last ten years) annual accounting return on assets (ROA), calculated in each
year as operating income (before amortization and depreciation and adjusted
for changes in the LIFO reserve) divided by the market value of the firm
in that year. l5 I measure performance over a ten-year period because the
evidence in Sweeney (1991) indicates that managers accounting procedure
choices are a function of firms long-term but not short-term earnings performance, which is consistent with there being relatively large costs of changing
accounting procedures. I subtract the change in the LIFO reserve to remove
the effect of inventory method choice from the accounting rate of return
variable.
A limitation of ROA is that because I measure performance over such a long
window, it may proxy for expected accounting returns as much as for poor
earnings realizations. Moreover, the ten-year average may not be a good proxy
for the relatively short periods of poor performance within this window that
could cause managers to switch to income-increasing accounting procedures. As
a result, I treat this variable as an (albeit noisy) control for both poor performance and expected returns, an approach that is justifiable on the grounds that it
is the other variables (the ios, size, and leverage) that are of primary economic
interest.
4.2. Descriptive statistics

Table 2 provides cross-sectional rank correlations for the independent variables described above. I report rank correlations because the distributions of
several variables are skewed.
The two financial leverage variables are inversely related (correlation - 0.63),
which is expected if both variables measure financial leverage. The ios variables
are also correlated, providing assurance that they are capturing the same
underlying construct. Asset beta is negatively related to the assets-in-place
measure ( - 0.24 with gross PP&E-to-value) and positively related to the R&D
variable (0.29) and q (0.40).Thus, firms with relatively more growth opportunities (and so relatively few assets-in-place) tend to have more systematic asset
risk. Gross PP&E-to-value is negatively related to the R&D measure ( - 0.10)
and to q ( - 0.61). These correlations are all statistically significant and in the
correct direction for these variables to be measuring assets-in-place vs. growth
I51 have also calculated
ROA by adding back interest expense
a.djustments, with similar results to those that I report below.
in addition
to these other
D.J.
422
Skinner,
Incestment
opportunities
und uccounting
choice
Table 2
Sp;arman
rank
_~~~_~~
correlations
for fiscal 1987 between the size, profitability,
financial
investment opportunity
set measures of 504 sample firms.. b
_~~~
- ~~~~~~
-~.-~ ~- --Financial leverage
_..._~~_~_ _~ .-~~
Accg
Size:
Value
opportunity
PPE/
value
Asset
beta
Debt/
assets
0.25
- 0.02
0.17
0.41
0.01
- 0.02
0.21
0.14
0.25
- 0.63
- 0.32
0.62
Financial
lererage:
Debt/assets
MVE,value
and
set variables
~~~~~~_~
ROA
Projitahiliry
Accg ROA
lnresrment
opportunit_v
Asset beta
PPE/value
R&D/sales
MVE,
value
Investment
leverage,
R&D/
sales
0.19
~_~~~~
Tobins
4
0.18
- 0.1 Id
- 0.04
0.21
- 0.38
- 0.24
0.22
- 0.19
0.56
- 0.24
0.29
- O.lOd
0.40
- 0.61
0.22
cariohles:
The numbers in the table are Spearman rank correlation

coefficients, calculated using firm level
data. There are 504 firms, although the correlations
are typically calculated with fewer observations
since most firms have missing data for at least one variable.
bThe variables are defined as follows. All variables (except asset beta, accounting
ROA. and
Tobins q) are calculated for each firm as the average over year-end 1985 through year-end 1987, and
so are based on three observations
for each firm.
Firm value
= market value of the firm (in millions of dollars)
market value of equity plus book value of debt;
Accg ROA
: ten-year average of the yearly ratio of operating
income (before depreciation
expense and after adjusting
for the change in the LIFO reserve) to firm value,
expressed as a percentage:
of long-term debt divided by the book value of total assets;
Debt/assets
= book-value
MVE/value
= market value of equity divided by the market value of the firm;
Asset beta
= beta of the firms stock multiplied
by the MVE!value
ratio;
PPE/value
= gross property, plant, and equipment (at historic cost) divided by the market value
of the firm:
R&D/sales
= research and development
expense divided by net sales, expressed as a percentage;
Tobins q
= estimate of Tobins q ratio (see text for details of calculations).
Significantly
dSiunificantly
P
different
different
from zero at the 0.01 level, two-tailed

from zero at the 0.05 level, two-tailed
test.
test.
opportunities. Finally, y is positively related to the R&D (0.22) measure, consistent with q measuring firms intangible investment opportunities.16
There is also evidence of an association between these firms investment
opportunities and their financial leverage. In particular, financial leverage is
1 also performed a factor analysis to try and gauge how many underlying or latent variables the
four ios variables were capturing. The factor analysis reveals that a single factor explains 49% of the
cross-sectional
variation
in these four variables, and that two factors together explain 73% of the
variation. Moreover,
when 1 cxcludc the risk measure, a single factor explains 55% of the variation
in the remaining
three variables. This evidence is consistent with these variables capturing
two
underlying investmcnl opportunity
constructs: the assets-in-place vs. growth options dimension, as
well as the fact that asset risk varies as a function of the ios.
positively related to the proportion of assets that comprise assets-in-place: the

equity-to-value measure, for example, is negatively related to gross PPE-tovalue ( - 0.38) and positively related to R&D/sales (0.22) and 4 (0.56). These
correlations are consistent with Myers prediction: firms with more assetsin-place are, on average, more highly levered than other firms. In addition, the
correlation between asset beta and equity-to-value is 0.62; so there is a negative
relation between leverage and systematic asset risk. This is to be expected if firms
with more growth opportunities are also more risky (i.e., because growth
opportunities represent more risky investment opportunities than assets-inplace, they also support less debt). Overall, the evidence confirms that financial
leverage is related to the nature of firms investment opportunities, which
supports the view that some of the extant evidence on the debt/equity hypothesis
could be confounded by an omitted variables problem.17
Firm size is related to three of the four ios measures. Size is positively related
to systematic asset risk (0.41), R&D/sales (0.19), and 4 (0.18). The accounting
ROA variable is positively related to asset beta (0.14) and negatively related to
market-valued leverage (0.2 1 with equity-to-value).
5. The relation between the investment opportunity set and firm contracts
This section provides evidence on the relation between the ios and nature of
sample fi*?ms management compensation plans and debt contracts. This linkage
is important because the arguments in section 2 suggest that the ios affects
accounting choice indirectly through its effects on the nature of firms contracts.
I collect information
on these firms bonus plans on two levels of detail.*

First, to investigate whether there is a relation between the parameters of these
plans [such as the lower and upper bounds that Healy (1985) describes] and the
ios, I collect all available plan details for a subsample of 100 sample firms.
Second, consistent with previous studies, I also collect information on whether
or not firms in the full sample had a bonus plan under which the bonus award
explicitly dep=~$5 Oii accounting earnings.
Larger firms are more likely to disclose bonus plan details in their proxy
statements than are smaller firms [see Healy (1985), Kim and Schroeder (199O)J.
To maximize the chances of obtaining bonus plan details, I collect all available
bonus plan information for the 100 !argest sample firms from the proxy
See Bradley, Jarrell. and Kim (1984). Gaver and Caver (1993). Long and Mak
Smith and Watts (1991) for similar evidence.
Following
PliltIS.
Heidy (1085) and other subsequent slttdics I restrict attetiiion
f 1985). and
to firms annual bonus
424
D.J. Skinner, Inaestment opportunities and accounting choice
statements that these firms filed over the 1981 through 1990 period. (By selecting
only the largest firms I weaken the power of these tests if, as the correlations in
table 2 suggest, size is correlated with the ios: by choosing large firms I tend to
select firms with relatively more assets-in-place than characterizes the sample as
a whole.) I chose the ten-year sampling interval because previous studies report
that bonus plans are subject to shareholder approval every three, five, or ten
years, at which time details of the plans are usually available in the proxy
statement [e.g., see Healy (1985)].
Of these 100 firms, 9 appear not to have a bonus plan, 6 disclose the existence
of a plan but provide no information about how awards are determined under
the plan, 44 have a discretionary bonus plan (under which the bonus award is
made at the discretion of the compensation committee of the board of directors),
and 41 have incentive bonus plans that explicitly tie the bonus to an accounting
earnings measure (return on equity, operating earnings, net income before taxes,
extraordinary items, etc.).l9 These proportions are comparable to those in Healy
(1985). Healy starts with the Fortune 250, and finds that around half of these
firms (123) have bonus plans but do not publicly disclose bonus plan details, and
ultimately arrives at a sample of 94 firms with bonus plans that tie the bonus to
accounting earnings. These proportions are similar to mine because a large
number of firms in the former group probably have discretionary bonus plans
that Healy excludes.
To investigate whether there is a relation between the ios and the nature of
firms bonus plans, table 3 compares the size, leverage, profitability, and ios
measures for those firms: (i) with incentive bonus plans (41 firms), (ii) with
discretionary bonus plans (44 firms), and (iii) without any bonus plans (9 firms).
If the arguments in section 2 are correct, accounting earnings should be
relatively better performance measures in firms with more assets-in-place.
The terms incentive and discretionary bonus plans are from Sibson (1990, ch. 18). Although
a detailed examination of this issue is best left for further research, my reading of the available details
of these 85 bonus plans indicates clearly that two distinct types of plan exist, where one type
(discretionary) ostensibly provides boards with much more discretion than the other. More
specifically, incentive bonus plans typically specify - with accounting e;mings numbers - the upper
and lower bounds, and then give the compensation committee discretion to decide on the bonus
within these bounds. For example, the General Motors Bonus Plan provides for a transfer to the
reserve of 8% of net earnings in excess of $I billion, provided that amount does not exceed the total
amount of the common stock dividend for the year. In contrast, discretionary plans provide the
compensation committee with almost complete discretion with respect to the bonus. For example,
Union Carbides 1989 proxy statement, in describing the bonus plan, states that the bonuses, if any,
for 1988 will be determined by the Committee. . . on the basis of the Committees evaluation of
factors such as the Corporations financial results . . . , the performance of the officers individually
and . . . as a group, and the levels of salaries and bonuses paid by competitive employers. The
Committee has the right, in its sole discretion, to increase, decrease, or eliminate the bonus
the basis of the Committees review (Union Carbide proxy statement dated March 17. 1988,~.2~
The existence of such a dichotomy makes sense if accounting earnings are better measures of firm
performance in some firms than they are in others, as 1 argue below.
D. J. Skinner, Investment opportunities and accounting choice
425
Table 3
Comparison of average (median) size, leverage, and investment opportunity set measures for 94
sample firms: (i) with an accouniing-earnings-based incentive bonus plan. (ii) with a discretionary
bonus plan that is not explicitly tied to accounting earnings, and (iii) without any type of bonus plan,
in fiscal 1987.
Firms with a
discretionary
bonus plan
(44)
Firms with an
incentive
bonus plan
(41)
11,747
(11,187)
7,407
(6,638)
14,184
(6,791)
Debt/assets
0.148
(0.106)
0.177
(0.143)
0.207
(0.213)
MVE/value
0.717
(0.809)
0.648
(0.680)
0.608
(0.602)
0.115
(0.109)b
0.124
(0.119)
0.131
(0.134)
Asset beta
0.824b
(0.862)
0.706
(0.705)
0.679
(0.670)
PPE/value
0.474
(0.266)
0.394
(0.344)
0.526
(0.461)
R&D/sales
0.054
(0.075)b
0.042b
(0.023)b
0.026
(0.012)
Tobins q
1.51
(1.52)
1.36
(1.20)
1.25
(1.16)
Firms without
a bonus plan
(9)
Firm size:
Value
Financial leverage:
Projtability:
Accounting ROA
Investment opportunity variables:
The variables are defined as follows. All variables (except asset beta, accounting ROA, and
so are based on three observations for each firm.
Value
= market value of equity plus book value of debt;
Debt/assets = book-value of long-term debt divided by the book value of total assets;
MVE/value = market value of equity divided by the market value of the firm;
Asset beta = beta of the firms stock multiplied by the MVE/value ratio;
PPE/value
= gross property, plant, and equipment (at historic cost) divided by the market value
of the firm;
R&D/sales = research and development expense divided by net sales, expressed as a percentage;
Tobins q
bStatistically significant difference between the number in this column and the corresponding
number for firms with an incentive bonus plan at the 10% level under two-tailed, two-sample f-tests
(Wilcoxon tests).
Statistically significant difference between the number in this column and the corresponding
number for firms with an incentive bonus plan at the 5% level under two-tailed, two-sample r-tests
(Wilcoxon tests).
426
Consequently, these firms are likely to use incentive bonus plans that tie the
bonus directly to accounting earnings while firms with more growth opportunities are likely to use discretionary bonus plans that do not.
The evidence in table 3 is consistent with this. Firms that use incentive
bonus plans have higher mean and median gross PP&E to value ratios
(more assets-in-place) but smaller q and R&D ratios (fewer growth opportunities) than firms that use discretionary bonus plans, although only the difference in mean and median R&D ratios is statistically significant. Moreover,
firms without any bonus plans appear to be those with the largest proportion
of growth opportunities of all of the three groups, which is also consistent
with accounting numbers being poorer performance measures in growth
firms.
The evidence in table 3 provides some support for the idea that the ios affects
the firms choice of bonus plan (i.e., incentive vs. discretionary) such that firms
with relatively more assets-in-place are more likely to use incentive bonus plans.
If this is the case in general and other things are held the same, managers of these
firms are likely to have larger incentives to choose income-increasing accounting
procedures, implying an indirect link between the ios and accounting choice.
This is examined further in section 6.
For those firms with an incentive bonus plan that ties the bonus to an
accounting earnings measure (41 firms), I also examine whether there is a relation between how the parameters in these firms plans are set and their leverage
and ios measures. Of these 41 firms, 26 disclosed details of a lower bound and 16
disclosed details of an upper bound. The tests (not reported in tables) indicate
that there is no apparent relation between these bounds and the ios or leverage
of these firms. However, this is not too surprising since: (i) the results in table
3 indicate that firms with incentive bonus plans tend to be those with relatively
large amounts of assets-in-place, so there is not much variation in the ios of the
firms that disclose details of the upper and lower bounds, and (ii) the sample
sizes are small.
5.2.
The ios and accounting-based debt conuenunts
I collect data
on sample firms accounting-based

debt convenants from
Moodys Industrial Marum (1987). Begley (1990b) indicates that Moodys provides reasonably complete data on accounting-based debt covenants in firms
public but not private debt issues. As a result of this, I can only make inferences
about the relation between the ios and accounting-based debt covenants in
public debt issues. This limitation will be important if there is any systematic
relation between the ios and firms relative utilization of public vs. private debt,
D. J. Skinner,
Incesrnzenr opporrunities
und accounting
choice
427
as seems likely. 2o Following Duke and Hunt (1990) and Press and Weintrop
(1990), I collect information on the existence of nine specific types of accountingbased covenants, although these nine covenants can be classified into the four
broad groups that both of these papers use: leverage restrictions, working
capital restrictions, net asset restrictions, and retained earnings restrictions.
Of the 504 sample firms, I am able to identify 465 in Moodys. Of these firms,
259 (56%) have at least one accounting-based debt covenant, while the remaining 206 do not. To examine the relation between accounting-based debt covenants, financial leverage, and the ios, table 4 provides a comparison of the firms
with at least one accounting-based
debt covenant to the firms without an
accounting-based debt covenant. 21 There are two principal results in the table.
First, consistent with the evidence in Begley (1990a) and Duke and Hunt (1990),
firms with accounting-based debt covenants have more debt than other firms.
The average (median) debt-to-assets ratio for the firms with at least one accounting-based debt covenant is 0.25 (0.23) compared to an average (median) of 0.16
(0.14) for the firms with no accounting-based debt covenants. These differences
are significant at the 1% level. Similar differences exist for the equity-to-value
leverage measure. Thus, firms with more debt are more likely to use accountingbased debt covenants in public debt agreements, as argued in section 2.
Second, there are also reliable differences between the ios measures for firms
with and without accounting-based debt covenants. On average, the firms with
accounting-based debt covenants in their public debt agreements have lower
asset betas (0.59 vs. 0.67), higher gross PP&E ratios (0.59 vs. 0.51), and smaller
R&D (0.018 vs. 0.028) and q (1.09 vs. 1.29) ratios than the firms without
accounting-based
covenants. Differences between the medians are of similar
magnitude, and all of these differences are statistically significant at the 5% level
for two-tailed, two-sample t and Wilcoxon tests. This evidence is consistent with
the prediction from section 2 that firms with more assets-in-place have more
debt, and so are more likely to use accounting-based debt covenants than other
firms [although again, this relation holds only for public debt - growth firms
2DAsdiscussed earlier, it is likely that managers in growth firms have more specific knowledge
about their firms assets and that the value of these growth opportunities is less observable. There is
therefore likely to be a relatively large informational asymmetry between managers in growth firms
and outsiders, especially if part of asset value comprises information that is proprietary. If this is so,
then these firms may favor the use of private debt over public debt for two reasons. First, contracting
and negotiation costs will be lower, because it is probably easier to convince a few relatively
sophisticated lenders of the value of their firms assets. Second, it is less likely that the proprietary
information will end up in the public domain if the firm negotiates with private lenders.
210f the firms with at least one accounting-based convenant, 69 had one covenant, 90 had two
covenants, 83 had three covenants, 14 had four covenants, and 3 had five covenants. When
I calculate the rank correlations between the number of accounting-based covenants and the
leverage and ios measures, I obtain similar inferences to those described in the text. That is, there is
a positive relation between the number of accounting-based debt covena?& financial leverage. and
assets-in-place.
D.J.
428
Skinner,
Inoestment
opportunities
and accounting
choice
Table 4
Comparison of average (median) size, leverage, and investment opportunity set measures for 465
firms with and without at least one accounting-based debt covenant in their public debt agreements
in fiscal 1987.eb
Firms with
accounting-based
debt covenants
(259)
Firm
Firms without
accounting-based
debt covenants
(206)
Test statistic
for
difference
size:
1,902
(656)
4,134
(616)
t = 2.84*
z = 0.21
Debt/assets
0.251
(0.227)
0.164
(0.138)
t = - 6.22*
Z = - 6.71*
MVE/value
0.521
(0.532)
0.654
(0.671)
t = 7.51*
Z = 6.99d
0.119
(0.123)
0.120
(0.126)
t = 0.22
z = 0.43
Asset beta
0.587
(0.597)
0.674
(0.663)
t = 2.71*
Z = 2.47*
PPE/value
0.588
(0.507)
0.508
(0.407)
t = - 1.95
Z = - 3.46*
R&D/sales
0.018
(0.002)
0.028
(0.007)
t = 2.95d
Z = 2.37d
Tobins q
1.09
(0.97)
1.29
(1.11)
t = 3.29d
z = 3.43*
Value
Finanrial
leverage:
Profitability:
Accounting ROA
Investment
opportunity
variables:
The numbers in the difference column are test statistics from two-sample t-tests (Wilcoxon tests)
of the null hypothesis that the average (median) difference between the two groups of firms is zero.
bThe variables are defined as follows. All variables (except asset beta, accounting ROA, and
so are based on three observations for each firm.
Value
market value of equity plus book value of debt;
Debt/assets 1 book value of long-term debt divided by the book value of total assets;
MVE/value = market value of equity divided by the market value of the firm;
Asset beta = beta of the firms stock multiplied by the MVE/value ratio;
PPE/value = gross property, plant, and equipment (at historic cost) divided by the market value
of the firm;
R&D/sales = research and development expense divided by net sales, expressed as a percentage;
Tobins q
Statistical significance at the 5% level (two-tailed test).
*Statistical significance at the I % level (two-tailed test).
D. J. Skinner, Incestntent opportunities and accounting choice
429
may use more private debt which also uses accounting-based debt covenants
- see Leftwich (1983) 3. Thus, the evidence suggests that the ios drives both
financiai leverage and the nature of debt coniracts, which implies that there is
likely to be an indirect link between the ios and accounting choice, as outlined in
section 2. I address this directly next.
6. The relation between accounting procedure choices and size, leverage,

and the investment opportunity set
6.1. The accounting procedure choices
This section provides evidence on the determinants of three accounting

procedure choices: inventory cost flow assumption, depreciation method, and
goodwill amortization period. To aggregate these choices across firms, I develop
a scale (from 0 to 2) to measure the extent to which a given accounting
choice can be characterized as income-increasing. Specifically, I characterize
the choice of first-in-first-out
(FIFO) inventory cost flow assumption, of
straight-line depreciation, and of a 40-year goodwill amortization
period
(the maximum) as income-increasing; these choices are assigned a score of 2.
Conversely, the use of a last-in-first-out (LIFO) cost flow assumption, accelerated depreciation, and a goodwill amortization period of less than 30 years
are characterized as income-decreasing and assigned a score of 0. Finally,
the average cost method for inventory (or an approximately equal-weighted
combination
of FIFO and LIFO), units-of-production
depreciation (or a
combination of straight-line and accelerated methods), and a goodwill amortization period of between 30 and 39 years (or a statement that goodwill is
amortized over periods not to exceed 40 years) are characterized as neither
income-increasing nor income-decreasing, and so receive a score of 1. Therefore,
for each of the three procedure choices, the score increases from 0 for the most
income-decreasing technique through to 2 for the most income-increasing
technique.22
The characterization of accounting techniques as income-increasing or income-decreasing is a simplification. The choice of particular accounting techniques is a choice that determines how revenues and expenses are allocated to
different accounting periods rather than a choice that affects the total amount of
income that is recognized. However, without access to additional information it
is difficult to improve on the above categorization. Moreover, since previous
studies also adopt this categorization scheme [e.g., Hagerman and Zmijewski
(1979), Zmijewski and Hagerman (198 I), Dhaliwal, Salamon, and Smith (1983,
221 alsa use finer partition where suffkient detail is provided in the annual report. For example, if
a firm states that it uses a combination
oT the FIFO and AC methods, I code this choice as 1.5.
430
D.J. Skinner,
~~l~.~t~lle~ll
oppomrrilies
urtd
uCCouFltiFlg
CilOiCe
Sweeney (1991)], my results will be directly comparable to those already in the

literature.
Another difficulty with the way that I characterize these firms accounting
choices is that it ignores the higher moments of the earnings effects of different
accounting methods. For example, LIFO will often increase the variability of
firm earnings vis-a-vis FIFO. This may be at least as important to risk-averse
managers as the effect of inventory method on the level of earnings, especially
since the levels effects tend to be offsetting over time. Once again, however, it is
difficult to do a great deal about this without performing a much more detailed
time series analysis that considers more precisely the effects of alternative
accounting methods on the variance of individual firm earnings.
Table 5 provides a summary of these firms accounting choices. With respect
to inventory accounting, the majority (38%) of firms choose LIFO, while 32.5%
choose FIFO, and 20.7% make an intermediate choice. The proportion of firms
using LIFO is comparable to that in Accounting Trends and Techniques 1988
(ATT) of 40%. The depreciation method choice shows less variation, with
74. I % of the sample choosing straight-line depreciation, 14.9% choosing either
units-of-production
or an intermediate combination, and only 9.6% choosing
accelerated depreciation. 23 The fact that 74% of firms make the incomeincreasing depreciation choice, while only 32% of firms choose the incomeincreasing inventory method suggests that different forces drive these two
accounting choices; in particular, the inventory choice likely reflects tax-minimization as well as financial reporting incentives. Nevertheless, the depreciation
and inventory choices exhibit a small positive correlation of 0.10 (significant at
the 5% level). Finally, of the 231 firms that report a goodwill accounting
method, around half (48%) make the income-increasing choice, 40.7% make an
intermediate choice, leaving 11.3% that choose a relatively small amortization
period. The goodwill choice exhibits a small positive correlation with the
depreciation choice (correlation of 0.10) but is uncorrelated with the inventory
choice (correlation of 0.02).
The next subsection provides evidence on the determinants of these firms
inventory, depreciation, and goodwill choices. I analyze these choices separately
because the relatively small correlations among them indicates that: (i) different
forces drive the different accounting choices (e.g., tax considerations affect firms
inventory accounting choices but not their depreciation or goodwill choices),
and/or (ii) the noise added by categorizing particular choices as income-increasing or income-decreasing
(see above) limits my ability to aggregate these scores
across the three different accounting choices.
230f the 600 firms in the ATT sample, 559 use straight-line
depreciation,
132 use accelerated
methods, and 51 use units of production (clearly, a number of firms use more than one mclhod).
Similarly, of the 213 firms that report a goodwill amortization
period, 173 (HI%) choose 40 years
(and another 77 disclose that they use ;I period not exceeding 40 years).
D. J. Skitmer,
hw~rtmw
opporrutriries
unci accomtit~g
choice
431
Table 5
Summa ry of the inventory, depreciation. and goodwill accounting choices of 504 sample firms in
fiscal 1987.
_____
Choice
Code
Number
Percent of totaf
(A) Itmwtory
FIFO
AC,FIFO
AC or Mix
AC,LIFO
LIFO
cost jlow
assutnptiot~
2
1.5
1
0.5
0
152
26
97
15
178
Total
468
_______
(B) Depreciation
SL
UPiSL
UP or Mix
U P/Acc.
Act.
Total
__.__~~__
method
2
1.5
A.5
369
7
74
0
48
years
30-39 years
< 30 years
2
:,
40
Total
74.1
1.4
14.9
0.0
9.6
498
~~ _~~~~~_~ ______~_._
(C) Goodwill
attlortizatiot~
period
111
94
26
48.0
40.7
11.3
231
D~fitzition
FIFO
LIFO
AC
Mix
SL
Act.
UP
Mix
32.5
5.4
20.7
3.2
38.0
=
=
=
=
=
=
=
=
of acron~ws:
first-in-first-out inventory cost flow assumption;

last-in-first-out inventory cost flow assumption;
average cost method;
approximately equal-weighted combination of FIFO and LIFO;
straight-line depreciation;
accelerated depreciation method;
units-of-production depreciation method;
combination of SL and Act. methods.
6.2. Multivariate tests

Tables 6 and 7 provide maximum likelihood estimates of ordinal-level logit
models of the inventory, depreciation, goodwill, and combined depreciation-goodwill accounting choices. 24 In each case the dependent variable is
coded as described above (table 5) so that a positive coefficient indicates an
24Maddala (1983, pp. 46-49) discusses multinomial models with ordered responses and Cox and
Snell(1989, ch. 5) describe the ordinal-level logit model that I use here. This type of procedure is also
used by Zmijewski and Hagerman (19Pl) and is appropriate given the ordinal nature of the
accounting choice data. I have estimated a number of the equations reported in tables 6 and 7 using
the ordinal-level probit (or n-probit) model, with very similar results. The ordinal-level logit model is
implemented using the MIDAS statistical package developed by the Center for Statistical Consulting and Research at the University of Michigan. This package uses the Newton-Raphson numerical
procedure to iteratively obtain estimates of the parameters that maximize the likelihood function.
The. pseudo R-square that I report is from Aldrich and Nelson (1984, p. 57).
432
D.J. Skinner,
Incesstntenr opporlunities
and accounting
choice
increase in the probability of choosing a more income-increasing accounting

alternative. To investigate whether the ios variables help explain accounting
choice, I report p-values from chi-square tests of the joint null hypothesis that all
of the coefficients on the ios variables are equal to zero. This test examines
whether or not the ios variables, as a group, add statistically significant explanatory power to the regression. I also report (in parentheses beneath each estimated coefficient) the coefficient divided by its asymptotic standard error and
refer to this ratio as a t-statistic, as well as an overall regression p-value and
pseudo R-square.
Because of the large number of independent variables in these regressions (I
include the traditional size, financial leverage, and compensation variables,
along with ROA and the four ios proxy variables), there are many possible
regression specifications. To economize on space, I report on only a few of these
regressions in tables. However, because the results (i.e., which variables are
significant and which are not) do not vary greatly between different specifications, I am able to do this without distorting the impressions one obtains from
the data. In addition, in all of the tables I report on the regression that includes
all of the independent variables. (A more comprehensive set of regression results
is available from the author.)
Another issue with respect to model specification is whether there is an
important simultaneity problem here. If it is the case (as I discuss in section 2)
that the ios determines firms financing, compensation, and accounting choices,
then the error terms in these regressions will be correlated with some of the
independent variables, leading to inconsistent estimates of the coefficients.
However, to estimate a simultaneous equation model requires specifying exactly
what the relation between the ios and these other variables is, which is difficult
given the limited state of our knowledge in this area [see Smith and Watts
(1991)].
Table 6 provides the results for the inventory method choice regressions. The
first equation provides evidence on the size, bonus plan, and debt/equity
hypotheses that other studies investigate. Consistent with the size hypothesis,
the coefficient on firm size is reliably negative (t-statistic of - 5.35). However,
the coefficients on financial leverage and the bonus plan dummy are not
significant, which is inconsistent with the debt/equity and bonus plan hypotheses. In the second regression I include the ROA measure along with firm size and
financial leverage. The coefficient on this variable is reliably negative, which is
consistent with firms with poor accounting performance making income-increasing accounting choices, although the negative coefficient may also indicate
that riskier firms are more likely to choose LIFO.
Regression 3 in table 6 includes the four ios variables, I am able to reject the
null hypothesis that the coefficients on these variables are all zero at the 0.0001
level. The pseudo R-square for this regression is 6.1 percent. The coefficient on
the PP&E variable (which increases with assets-in-place) is reliably negative,
while that on R&D is reliably positive, implying that firms with relatively more
assets-in-place are more likely to make the income-decreasing
accounting
choice.
Regression 4 in tab!e 6 includes firm size along with the four ios measures as
explanatory variables. Including the ios measures does not reduce the statistical
significance of the size relation, and conversely, the statistical significance of the
ios variables is not reduced by including firm size in regression. The p-value for
the joint null hypothesis that the coefficients on the ios variables are all zero is
significant at the 1% level and the pseudo R-square - of 14.3% - is larger than
for the regressions that include either just the size variable or just the ios
variables. It is again the case that the coefficient on the PP&E measure is
reliably negative while that on R&D is reliably positive. The inferences do not
change a great deal when I include all of the independent variables in regression
5. In addition to the size and ios variables, the ROA variable is negative and
significant, and the pseudo R-square increases to 15.4%. Overall, this evidence
indicates that the ios variables add explanatory power to the inventory regressions over and above that provided by firm size and ROA.
These results suggest that firms with relatively more assets-in-place are more
likely to choose LIFO, which I have characterized as the income-decreasing
accounting choice. However, the FIFO vs. LIFO choice also depends on these
firms relative tax positions and on the relative tax benefits of these two
accounting choices. The relative tax benefits of LIFO depend on the direction
and rate of change in firms input prices, on the variability of firms physical
inventory levels, and on firms effective tax rates [e.g., see Lee and Hsieh
(1985) and Dopuch and Pincus (1988)]. These characteristics are likely to be
correlated with the ios because: (i) the nature of the firms production process is
likely to affect the variability of inventory levels, and (ii) the firms production
process determines the inputs that are required. The more variable the firms
inventory levels, the higher the probability of forced LIFO liquidations, and SO
the less attractive is LIFO. Thus, firms with relatively more growth opportunities, because they also tend to be more risky, are less likely to select LIFO
than are other firms. So the tax effects of the firms ios generate a prediction
that firms with more assets-in-place will make the income-decreasing accounting choice.
To investigate the tax explanation, I include in regressions 6-10 in table
6 a variable (Infl.) that measures the direction and rate of price change in these
firms industries. Other things equal, the relative tax benefits of LIFO increase
with higher (more positive) rates of change in firms input prices. I use data from
the Bureau of Labor Statistics Producer Price Indices to group these firms
industries into three equal-sized groups: those with high, medium, and low rates
of price change since 1967 (similar to other studies, these are industry-level
output price indices). I then define the Infl. variable such that firms in the low
group are coded 2, those in the medium group are coded 1, and those in the high
D.J. Skinner, hestment
opportunities and accounting choice
0.36
(0.41)
- 0.26
( - 3.36)
10.
- 6.23
0.35
(1.60) ( - 2.02)
0.32
(1.50)
0.01
(0.02)
_~ ~~__
- 0.12
( - 0.25)
- 0.58
( - 1.19)
- 0.73
( - 1.54)
0.24
(0.80)
0.30
(0.99)
8.76
(2.17)
6.09
(1.44)
0.0800
0.0169
0.0000
0.0000
0.163
0.15 1
=
=
=
=
opportunity
set (10s) measures:
beta of the firms stock multiplied by the MVE/value ratio:

gross property, plant, and equipment (at historic cost) divided by the market value of the firm;
research and development expense divided by net sales, expressed as a percentage;
estimate of Tobins q ratio (see text for details of calculations).
The ios p-value is that from a test of the joint hypothesis that the coefficients on the ios proxy variables are all equal to zero. More specifically, the
ditference between minus twice the log-likelihood for the full model (including the ios variables) and minus twice the log-likelihood for the reduced model
(excluding the ios variables) follows an asymptotic chi-square distribution with degrees of freedom equal to the number of ios variables. See, e.g., Aldrich
and Nelson (1984, p. 59).
Beta
PPE
R&D
q
insestment
The table reports on multinomial logit regressions of firms inventory accounting scores (coded from 5 for FIFO through 1 for LIFO) on various
combinations of the following explanatory variables.
bAll explanatory variables (except beta, ROA. and the dummy variables) are calculated for each firm as the average over year-end 1985 through
year-end 1987 values, and so are based on three observations for each firm.
Value
Debtassets
= book value of long-term debt divided by the book value of total assets;
D:A*covenant = debt,assets multiplied by a dummy variable set equal to one if the firm has at least one accounting-based debt covenant and zero
otherwise;
= dummy variable set to one if the firm has a bonus plan where the amount of the award is explicitly tied to an accounting earnings
Bonus
number and zero otherwise:
= ten-year average of the yearly ratio of operating income (before depreciation expense and after adjusting for the change in the LIFO
ROA
reserve) to firm value. expressed as a percentage;
= variable coded as 2/1/O for firms in industries that experienced low/medium/high rates of price change since 1967, as measured by the
Infl.
Bureau of Labor Statistics Producrr Price Indices.
0.47
(0.53)
- 0.31
( - 4.07)
9.
436
D.J. Skinner,
Jncestment
opportunities
and accounting
choice
group are coded 0, If the relative tax benefits of LIFO increase with higher rates
of increase in firms input prices (and this helps explain the accounting choice),
the Infl. variable will be positively related to accounting choice.25
The evidence indicates that tax-related incentives do help explain firms
inventory method choices. When I include the Ink variable with firm size and
financial leverage in regression 6, its coefficient is reliably positive (t-statistic of
3.77), consistent with the prediction that the relative tax benefits of LIFO
increase with increases in input prices. The coefficient on firm size remains
reliably negative. Similarly, when I include the Infl. variable along with firm size,
financial leverage, and the ROA measure in regression 7, the coefficients on firm
size and ROA are reliably negative and that of InIX reliably positive.26 [Some
care must be exercised in comparing the results in equations 6-10 with those in
equations l-5 since there are 357 firms with available data for equations l-5 but
only 256 firms with available data for equations 6-10 (i.e., there are 101 firms
without the inflation data).]
Regressions 9 and 10 in table 6 include the ios variables along with firm size,
ROA, and the InP.. measure. The ios measures are less significant when all of
these other variables are included than they were in regressions l-5, suggesting
that the ios is correlated with tax incentives. (The fact that the ios variables are
jointly significant at the 0.0002 level in regression 8 indicates that the different
sample is not the cause of the decline in significance for these variables.)
Consistent with this, firms with relatively more growth opportunities tend to be
those for which the tax benefits of LIFO are relatively small: the InIl.variable is
negatively related to the PP&E variable (correlation of - 0.42) and positively
related to R&D (0.25) and 4 (0.25) so that firms with relatively more assetsin-place tend to have experienced larger rates of price increase.
2JI have also used this variable to perform a check on my coding of LIFO as income-decreasing
and FIFO as income-increasing. If input prices are falling over some period of time, then FIFO is
likely to be the income-decreasing accounting choice over that period. Therefore, it may be that my
coding is incorrect for firms for which the price variable is coded 2 (relatively low rates of price
increase) and correct for firms for which this variable is coded 0 (relatively high rates of price
increase). If this is the case, I would expect a different relation between the probability of choosing
FIFO and the independent variables for different levels of the price variable. However, when I run
these regressions with the price variable included as an interactive dummy with the other independent variables, none of the interactive terms are significantly different from zero at the 10% level,
which provides some assurance about the way that I code the inventory choice variable (these
regressions are not reported in tables),
261n unreported regressions I also include a measure ot :hese firms net operating loss (NOL)
carryforwards. Firms with relatively large NOL carryforwartis have little incentive to choose LIFO
to minimize taxes. Consistent with this argument, I find that firms that use FIFO have significantly
larger NOL carryforwards than firms that use LIFO (the mean and median differences are
significant at the 5% level). However, the NOL carryforward variable is not significant in a multivariate logit that also includes firm size, ROA, and lnfl. as independent variables. This could be due
to collinearity: not surprisingly, the NOL carryforward variable is negatively related to ROA
(correlation of - 0.44).
D. J. Skinner, hwestmenl opporrunitier and
accounting choice
437
Panel A of table 7 reports ehe results for the depreciation method choice. The
first regression in panel A characterizes extant results for the size, debt/equity,
and bonus plan hypotheses. Consistent with these hypotheses, the probability of
choosing the income-increasing accounting alternative (in this case straight-line
depreciation) increases as financial leverage increases, is larger when the firm has
a bonus plan that ties the bonus directly to accounting earnings, and decreases
as firm size increases. The coefficients on all three variables are reliably different
from zero, with t-statistics of - 3.72 (firm size), 4.09 (financial leverage), and 3.32
(bonus plan), and the pseudo R-square for this regression is 9.1 percent. The
second regression includes the ROA measure along with the size, leverage, and
bonus play variables. The coefficient on ROA is. insignificant.
According to the debt/equity hypothesis, financial leverage is important in
these regressions because it proxies for managers incentives to choose incomeincreasing accounting procedures to loosen accounting-based debt covenants. If
this is the case, there will be a relation between accounting choice and financial
leverage for firms with accounting-based covenants but not for firms without
such covenants. To test this, in regression 3 I include a dummy variable coded
1 if the firm has at least one accounting-based debt covenant and 0 otherwise,
multiplied by the debt/assets variable, as well as including the debt/assets
variable alone. The coefficient on this multiplicative dummy variable measures
how the slope coefficient on debt/assets differs for firms with and without
accounting-based debt covenants.
Similar to the results in Press and Weintrop (1990), the leverage variable
remains significant (t-statistic of 2.46) in this regression while the multiplicative
dummy variable is insignificant (t = 0.74). Thus, the relation between leverage
and the depreciation choice does not seem to be due to accounting-based
covenants, and so may reflect a direct effect of the ios on both accounting choice
and leverage. (However, this test is probably biased against the multiplicative
dummy variable because the debt/assets variable includes private as well as
public debt, while the debt covenant dummy includes only covenants in public
debt agreements.)
Regression 4 in panel A of table 7 includes the four ios measures. In this
regression the null hypothesis that all of the coefficients on the ios proxy
variables are zero is rejected at the 7% level. However, the only variable that is
individually significant is asset beta, which has a negative coefficient, implying
that firms with higher betas are more likely to select income-decreasing accounting procedures.
When I include the ios measures along with all of the other variables in
regression 5, they are only jointiy significant at the 10% level, and none of the
ios variables is individually significant. In contrast, the firm size, financial
leverage, and bonus plan variables remain statistically significant (their
t-statistics are largely unaffected by including the ios measures). Moreover, the
pseudo R-square in this regression is 9.8% which is similar to that for the first
438
D.J. Skinner, Inrestment
opportunities
and accounting
choice
regression in the table. The traditional variables thus seem to be more important than the ios variables. Because of collinearity in these data, this result does
not necessarily imply that size, leverage, and the bonus plan are the right
variables and that the ios measures are not. However, the fact that the
size, leverage, and bonus plan relations that other studies document are robust
to including the ios variables is important, because it suggests that these
variables do not matter in other studies merely because they are correlated
with the ios.
Table 7, panel B provides results for the choice of goodwill amortization
period. The results in the first regression are similar to, but weaker than, those
for depreciation: accounting choice is negatively related to firm size and positively related to both debt/assets and the existence of a bonus plan. However,
only the coefficient on debt/assets is reliably different from zero. The different
results may be because tlic sample of firms here is different to those in the
depreciation and inventory regressions: only 171 firms are included in panel
B compared with 369 firms in panel A and 357 firms in table 6. Moreover, the
171 firms included here tend to be larger than those in the depreciation
regressions (the difference in medians is significant at the 1% level) because the
smaller sample firms tend not to report a goodwill choice. When I reestimate the
depreciation regressions with only the 169 firms that report both goodwill and
depreciation choices, the size and bonus plan relations become insignificant.
In regression 2 of panel B I again include the multiplicative dummy variable
to see whether it is leverage per se,or the existence of accounting-based debt
covenants in combination with leverage, that better explains accounting choice.
In this regression the coefficient on the multiplicative dummy variable is positive
and significant at the 10% level, while the coefficient on debt/assets becomes
insignificant. Thus, for these firms leverage is only important for firms with
accounting-based covenants, which is consistent with managers selecting income-increasing accounting procedures to loosen debt-covenant constraints.27
Regression 3 includes the four ios measures as explanatory variabies. The ios
p-value is small (less than O.OOl), indicating that the ios variables together are
related to the goodwill choice, and the pseudo R-square is 10.2%, larger than for
either of the first two regressions in panel B. The two variables that are most
significant are PP&E (t-statistic of - 1.75) and R&D (t-statistic of - 3.45).
While the negative coefficient on R&D suggests that firms with more growth
opportunities are more likely to select income-decreasing accounting procedures, the negative coefficient on gross PP&E suggests the opposite.
*These results are different to those in panel A: the multiplicative dummy variable was not
significant in the depreciation regressions, while the debt/assets variable was significant. This
difference is not due to the different samples - when I reestimate the depreciation regressions with
only the 171 firms included in panel B, the coefficient on the debt/assets variable is still reliably
positive, while that on the multiplicative dummy is still insignificant.
D. J. Skinner, hcesmrenr
opportunities and uccounting choice
439
Regressions 4 and 5 in panel B of table 7 include the firm size, financial

leverage, bonus plan, and ROA measures along with the ios variables. The ios
variables, as a group, remain significant in the presence of the other variables.
Moreover, the coefficients on PP&E and R&D are both negative, although only
the coefficient on R&D is statistically significant. In addition, the coefficient on
the multiplicative leverage variable is positive and significant at the 5% level,
while the coefficient on the bonus plan dummy is positive and signi5cant at the
10% level. The results for goodwill are therefore consistent with the debt/equity
and bonus plan hypotheses, but also indicate that the ios variables are important.
Finally, table 7, panel C reports results for regressions that combine these
firms depreciation and goodwill choices in a single dependent variable, similar
to Zmijewski and Hagerman (1981). 28 For these regressions I categorize the
accounting choices of these firms into five different strategies, from most income-increasing (where the firm selects the most income-increasing alternative
for both depreciation and goodwill) to most income-decreasing (where the firm
selects the most income-decreasing alternative for both choices), and use these
strategies as the dependent variable. 29 As with the Zmijewski-Hagerman
tests,
this should provide a more powerful test than considering either of these choices
in isolation.
The evidence in the first two regressions in panel C of table 7 is again
consistent with the traditional contracting explanations for accounting choice.
The coefficients on the size, financial leverage, and the bonus plan dummy
variables are all reliably different from zero, with the size variable negatively
related to accounting choice and the debt/assets and bonus plan dummy
variables both positively related to accounting choice. Regression 3 includes
these variables along with the multiplicative debt covenant variable. The coefficient on this variable is positive and marginally significant (t = 1.53), supporting
the interpretation that leverage is important because it proxies for the existence
of accounting-based debt covenants (again, the coefficient on this variable may
be downward biased because the debt covenant dummy systematically excludes
covenants from private debt agreements). This regression has a pseudo R-square
of 15.4%.
Regression 3 in panel C includes the four ios variables. These variables are
jointly significant at the 1% level, and the regression R-square is 10%. The
1 do not include the inventory choice because, as noted above, this choice depends on tax as well
as contracting considerations, and so seems to be a different decision from the other two accounting
choices.
29Since the goodwill choice has only three levels (income-increasing, intermediate, and incomedecreasing), I also collapse the depreciation choice into three levels. Since there are three levels for
each of two choices, there are nine possible permutations and five unique strategies. Of the 229 firms
that report both choices, 85 (37%) select strategy 5 (the most income-increasing strategy), 85 (37%)
select strategy 4,39 (17%) select strategy 3, 14 (6%) select strategy 2, and 6 (3%) select strategy 1 (the
most income-decreasing strategy).
- 0.18
- 0.10
- 0.17
( - 0.10)
- 0.03
( - 0.32)
( - 0.22)
5.
- 0.35
0.70
(0.49)
2.47
(2.35)
- 0.02
( - 0.23)
( - 1.05)
- 0.09
( - 1.16)
4.
3.
7
-.
1.
3.46
(2.55)
3.04
(2.14)
2.64
(1.93)
2.49
(1.87)
1.17
(0.96)
0.83
(3.19)
_.
3.83
(1.20)
1.99
(0.69)
0.43
0.56
(1.69)
0.62
(1.88)
( 1.64)
0.51
(1.40)
PP&E
( - 0.61)
- 0.18
- 0.15
( - 0.54)
R&D
- 17.25
(- 3.16)
0.05
- 1.15
(0.07) ( - 1.56)
- 5.25
( - 1.19)
- 17.45
( - 3.45)
- 18.83
( - 3.52)
- 1.17
( - 1.75)
2.81
(0.74)
- 7.87
( - 1.93)
- 0.56
( - 1.03)
- 0.31
( - 0.09)
(B) Goodwill rnerhodchoice (171 lirms)
( - 0.97)
- 0.44
( - 3.52)
- 0.26
0.88
(0.74)
0.83
(3.23)
0.85
(3.32)
Beta
____-
0.0001
0.0004
0.51
(i-81)
0.32
(1.04)
0.0007
0.0976
0.28
(0.96)
0.19
(0.74)
0.0703
ios
p-value
4
___~~__
- 0.02
( - 0.09)
10s meusures:
___~~__ __-
(A) Depreciation method choice (369 firms)
ROA
ROA:
5.
( - 2.97)
3.10
(2.46)
4.04
(4. IO)
- 0.25
( - 3.74)
4.05
(4.09)
- 0.23
( - 3.72)
Bonus
Bonus plan:
___
- 1.05
( - 2.67)
3.
7
_.
1.
D/A *
covenant
Leuerqe:
__-_-
Debt/assets
____
____
Value
~--
Size:
4.
0.0003
0.0001
0.0007
0.0091
0.0186
0.0000
0.0703
0.153
0.146
0.102
0.073
0.055
0.098
0.023
0.067
0.065
o.oooo
o.oooo
0.091
Pseudo
R-square
0.0000
Overall
model
p-value
Estimated coefficients (with asymptotic t-statistics in parentheses) from a multinomial logit model of the depreciation (panel A), goodwill (panel B), and
combined depreciation-goodwill
method choices of sample firms in fiscal 1987, estimation by maximum likelihood. A positive coefficient indicates
a higher probability of choosing a more income-increasing accounting method.a*b
Table 7
2.03
(1.19)
2.69
(1.85)
4.19
(3.86)
2.42
(1.75)
2.02
(1.53)
0.95
(2.97)
0.94
(3.06)
0.88
(2.87)
- I.17
( - 0.28)
- 2.69
( - 0.74)
- 0.40
( - 0.66)
- 1.11
( - 2.05)
- 1.23
( - 2.38)
- 0.98
( - 1.40)
- 0.63
( - 0.96)
- 0.75
( - 1.19)
0.31
(1.06)
0.18
(0.66)
- 13.9
( - 2.87)
- 10.4
( - 2.06)
0.18
(0.65)
- 13.2
( - 2.82)
choice (169 firms)
0.0012
0.0001
0.0008
0.0000
0.0020
0.192
0.101
0.100
0.154
o.oooo
0.0008
0.14s
o.oooo
opportunity set (IOS) measures:

beta of the firms stock multiplied by the MVE/value ratio;
gross property, plant, and equipment (at historic cost) divided by the market value of the firm;
research and development
expense divided by net sales, expressed as a percentage;
estimate of Tobins q ratio (see text for details of calculations).
The ios p-value is that from a test of the joint hypothesis that the coefficients on the ios proxy variables are all equal to zero. More specifically, the
difference between minus twice the log-likelihood
for the full model (including the ios variables) and minus twice the log-likelihood
for the reduced model
(excluding the ios variables) follows an asymptotic chi-square distribution
with degrees of freedom equal to the number of ios variables. See, e.g., Aldrich
and Nelson (1984, p. 59).
Investment
=
Beta
PPE
=
R&D =
=
4
=Panel A of the table reports on multinominal

logit regressions of firms depreciation
scores (coded from 5 for straight-line
depreciation
through 1 for
accelerated depreciation)
on various combinations
of the explanatory
variables. Panel B of the table reports on multinomial
logit regressions of firms
goodwill amortization
periods (coded from 3 for a 40-year period through 1 for a period of less than 30 years; sn various combinations
of the explanatory
variables.
Panel C of the table reports on multinomial
logit regressions
of firms combined
depreciation-goodwill
scores (coded from 5 for most
income-increasing
through 1 for most income-decreasing)
on various combinations
of the explanatory
variables.
bAli explanatory
variables (except for beta, ROA, and the two dummy variables) are calculated for each firm as the average over year-end 1985 through
year-end 1987 values, and so are based on three observations
for each firm.
Value
Debt/assets
= book value of long-term debt divided by the book value of total assets;
D/A*covenant
= debt/assets multiplied by a dummy variable set equal to one if the firm has at least one accounting-based
debt covenant and zero
otherwise;
Bonus
= dummy variable set to one if the firm has a bonus plan where the amount of the award is explicitly tied to an accounting
earnings
number and zero otherwise;
ROA
= ten-year average of the yearly ratio of operating income (before depreciation
expense and after adjusting for the change in the LIFO
reserve) to firm value, expressed as a percentage.
5.
4.
- 0.14
( - 1.58)
- 0.16
( - 1.92)
3
__
3.
- 0.17
( - 2.02)
1.
(C) Combined depreciation-goodwill
442
coefficients on R&D and beta are negative and reliably different from zero,
again suggesting that firms with more assets-in-place (fewer growth opportunities) are more likely to make the income-increasing accounting choice. In
regression 4 I include the ROA variable along with the ios variables. Similar to
the results in panels A and B, the ROA variable is not significant, and its
inclusion does not affect the significance of the ios variables.
When I include all of the variables in regression 5, the pseudo R-square
increases to 19.2% but only two of the coefficients are individually significant:
the coefficient on the bonus plan dummy is reliably positive and the coefficient
on R&D is reliably negative. Of the other variables, the one that is closest to
being significant at the 5% level is that on the multiplicative debt/assets
*covenants variable. The reduction in the individual significance of the other
variables probably reflects the collinearity that exists between these variables.
Overall, the evidence in table 7 suggests that the traditional explanations are
important, even after controlling for the effect of the ios. Nevertheless, the ios
variables do help explain these choices, both by themselves and in combination
with the traditional variables. The results in table 7 also suggest that firms with
more assets-in-place are more likely to choose income-increasing accounting
procedures, which is opposite to the result for the inventory choice.
7. Discussion and summary

This study provides evidence on the relation between firms investment
opportunities, the nature of their debt and compensation contracts, and their
financial leverage, size, firm performance, and accounting procedures. There are
three principal empirical results.
First, larger firms are more likely to select income-decreasing depreciation
and inventory procedures, more highly-levered firms are more likely to select
income-increasing depreciation and goodwill procedures, and firms with bonus
plans are more likely to select income-increasing depreciation and goodwill
procedures, than are other firms. These results confirm extant results in the
literature, that is, they are consistent with the size, debt/equity, and bonus plan
hypotheses. Importantly, the evidence indicates that these relations are robust
to including measures of the ios and/or accounting performance (ROA) in
accounting choice regressions.
Second, I document that the ios affects the nature of firm contracts: firms with
more assets-in-place are more likely to have accounting-based debt covenants in
their public debt agreements and are more likely to use bonus plans that tie the
bonus directly to accounting earnings. As a result, managers of firms with more
assets-in-place have larger incentives, given the nature of their firms contracts,
to select income-increasing accounting procedures. Therefore, the ios affects
accounting choice indirectly through its effect on the nature of firm contracts.
6). J. Skinner, Imesmzenr opporrunities und accounling choice
443
Third, the ios variables are correlated with all three of the accounting
procedures that I examine. Moreover, except in the case of depreciation, the ios
variables provide incremental explanatory power in regressions that also include
firm size, financial leverage, and the bonus plan dummy as independent variables. Thus, the evidence indicates that there is an association between the ios
and accounting procedure choice, even after controlling for managers contractual incentives to select particular accounting procedures.
More specifically, firms with larger ratios of assets-in-place to value are more
likely to make income-increasing depreciation and goodwill choices than other
firms. However, while the ios proxy variables are also related to these firms
inventory accounting choices, the directional association is opposite that for the
depreciation and goodwill choices: firms with larger ratios of assets-in-place to
value are more likely to make the income-decreasing accounting choice (LIFO)
than are other firms. One plausible explanation for this result is that firms
inventory method choices are also affected by the relative tax costs of LIFO vs.
FIFO, and that the relative tax costs of FIFO are smaller for firms with
relatively more growth opportunities.
Overall, the evidence in this paper increases our understanding of current
accounting practice in several ways. First, while there has been speculation in
the literature about the effect of the ios on accounting choice, this study is the
first to document this relation in a general way for a large, randomly-chosen,
sample of firms. Importantly, the evidence suggests that previous evidence on
the size, debt/equity, and bonus plan hypotheses cannot be explained by the fact
that those studies exclude the ios, which is a potentially important correlated
omitted variable. Second, this study documents that the ios systematically
affects the nature of firm contracts, and thus affects managers contractual
incentives to select particular accounting procedures. Finally, the evidence
provides some limited support for the view that the ios affects accounting choice
directly, which is expected if the ios affects which accounting procedures are
optimal ex ante.
References
AICPA, 1988, Accounting trends and techniques, 42nd ed. (AICPA. New York, NY).
Aldrich, John H. and Forrest D. Nelson, 1984, Linear probability, logit, and probit models (Sage
Publications, Beverly Hills, CA).
Ball, Ray and George Foster, 1982, Corporate financial reporting: A methodological review of
empirical research, Supplement to Journal of Accounting Research 20, 161-234.
Begley, Joy, 1990a, The use of debt covenants to control agency problems, Unpublished dissertation
(University of Rochester, Rochester, NY).
Begley, Joy, 1990b, Debt covenants and accounting choice, Journal of Accounting and Economics
12, 125-139.
Bradley, Michael, Gregg A. Jarrell, and E. Han Kim, 1984 On the existence of an optimal capital
structure: Theory and evidence, Journal of Finance 39, 8:7-880.
Clinch, Greg, 1991, Employee compensatton a,rcf fins resear-.-h and development activity, Journal
of Accounting Research 29. 59-78.
444
D.J. Skinner, Imestmcnt
opprtunities
and accounting
choice
Collins, Daniel W.. S.P. Kothari, Jay Shanken, and Richard G. Sloan, i992, Lack of !imeliness
versus noise as explanations for low contemporaneous returns<arnings relation, Working paper
(University of Rochester, Rochester, NY).
COX,
D.R. and E.J. Snell, 1989, Analysis of binary data, 2nd ed. (Chapman and Hall, London).
Dechow, Patricia M. and Richard G. Sloan, 1991, Executive incentives and the horizon problem: An
empirical investigation, Journal of Accounting and Economics 14, 51-89.
Demski, Joel S., James M. Patell, and Mark A. Wolfson, 1984, Decentralized choice of monitoring
systems, The Accounting Review 59, 16-34.
Dhaliwal, Dan S., Gerald Salamon, and E. Dan Smith, 1982, The effect of owner versus management
control on the choice of accounting methods, Journal of Accounting and Economics 4,41-53.
Dopuch, Nicholas and Morton Pincus, 1988, Evidence on the choice of inventory accounting
methods: LIFO versus FIFO, Journal of Accounting Research 26, 28-59.
Duke, Joanne C. and Herbert G. Hunt III, 1990, An empirical examination of debt covenant
restrictions and accounting-related debt proxies, Journal of Accounting and Economics 12,
45-63.
Foster, George, 1986, Financial statement analysis. 2nd ed. (Prentice-Hall, Englewood Cliffs, NJ).
Gaver, Jennifer J. and Kenneth M. Gaver, 1993, Additional evidence on the association between the
investment opportunity set and corporate financing, dividend and compensation policies, Journal of Accounting and Economics 16, forthcoming.
Hagerman, Robert L. and Mark E. Zmijewski, 1979, Some economic determinants of accounting
policy choice, Journal of Accounting and Economics 1, 141-161.
Hall, Bronwyn H., Clint Cummins, Elizabeth S. Laderman, and Joy Mundy, 1988, The R&D master
file documentation, Technical working paper no. 72 (National Bureau of Economic Research,
Cambridge, MA).
Hamada, Robert S., 1972, The effects of the firms capital structure on the systematic risk of common
stocks, Jottrnal of Finance 27,435452.
Healy, Paul M., 1985, The effect of bonus schemes on accounting decisions, Journal of Accounting
and Economics 7,85-107.
Holthausen, Robert W., 1990, Accounting method choice: Opportunistic behavior, efficient contracting, and information perspectives, Journal of Accounting and Economics 12, 207-218.
Holthausen, Robert W. and Richard W. Leftwich, 1983, The economic consequences of accounting
choice: Implications of costly contracting and monitoring, Journal of Accounting and Economics 5, 77- 117.
Kim, Kyungho, and Douglas A. Schroeder, i990, Analysts use of managerial bonus incentives in
forecasting earnings, Journal of Accounting and Economics 13. 3-23.
Lambert, Richard A. and David F. Larcker, 1987, An analysis of the use of accounting ;:nd market
measures of performance in executive compensation contracts, Journal of Accounting Research
25,85-125.
Lee, Chi-Wen Jevons and David A. Hsieh, 1985, Choice of inventory accounting meth,ods: Comparative analyses of alternative hypotheses, Journal of Accounting Research 23, 468-485.
Leftwich, Richard, 1983, Accounting information in private markets: Evidence from private lending
agreements, The Accounting Review 58, 23-42.
Leftwich, Richard, 1990, Aggregation of test statistics: Statistics vs. economics, Journal of Accounting and Economics 12, 37-44.
Lilien, Steven, Martin Mellman, and Victor Pastena, 1988, Accounting changes: Successful versus
unsuccessful firms, The Accounting Review 63, 642-656.
Lindenberg, Eric B. and Stephen A. Ross, 1981, Tobins r~ratio and industrial organization, Journal
of Business 54, l-32.
Long, Michael S. and Ileen B. Malitz, 1985, Investment patterns and financial leverage, in: B.
Friedman, ed., Corporate capital structure in the United States (University of Chicago Press,
Chicago, IL) 325-348.
Maddala, G.S., 1983, Limited-dependent and qualitative variables in econometrics, Econometrics
Society monograph (Cambridge University Press, Cambridge).
Merck, Randall, Andrei Shleifer, and Robert W. Vishny, 1988, Management ownership and market
valuation: An empirical analysis, Journal of Financial Economics 20, 293-315.
Myers, Stewart C., 1977, Determinants of corporate borrowing, Journal of Financial Economics 5,
147-175.
D. J. SliiiInei I .kccs:ment opporlunities and accounting choice
445
Press, Eric G. and Joseph B. Wcinrrop, ! 390, Accounting-based constraints in public and private
debt agreements: Their association with leverage and their
I, impact on accounting choice, Journal
of Accounting and Economics 12,65-95.
Sibson, Koberi E., i99g, Compensation, 5th ed. (American Management Association, New York,
NY)
Smith, Clifford W., Jr. and Jerold B. Warner, 1979, On financial contrac;ir~~*
.e. An analysis of bond
covenants, Journal of Financial Economics 7, 117-161.
Smith, Clifford W., Jr. and Ross L. Watts, 1982, Incentive and tax effects of executive compensation
plans, Australian Journal of Management 7, 139-157.
Smith, Clifford W., Jr. and Ross L. Watts, 1991, The investment opportunity set and corporate
policy choices, Working paper (University of Rochester, Rochester, NY).
Sweeney, Amy P., 1991, The impact of debt-covenant violations on accounting choices, Working
paper (Harvard Business School, Cambridge, MA).
Watts, Ross L.. 1977, Corporate financial statements, a product of the market and political
processes, Australian Journal of Management 2, 53-78.
Watts, Ross L. and Jerold L. Zimmermman, 1986, Positive accounting theory (Prentice-Hall,
Englewood Cliffs, NJ).
Watts, Ross L. and Jerold L. Zimmerman, 1990, Positive accounting theory: A ten year perspective,
The Accounting Review 65, 131-156.
Zimmer, Ian, 1986, Accounting for interest by real estate developers. Journal of Accounting and
Economics 8, 37-5 1.
Zmijewski, Mark E. and Robert L. Hagerman, 1981, An income strategy approach to the positive
theory of accounting standard setting/choice, Journal of Accounting and Economics 3,129-149.

1 s2.0 016541019390034D Main

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

1 s2.0 016541019390034D Main

Transféré par

Droits d'auteur :

Formats disponibles

Journal of Accounting and Economics 16 (1993) 407-445.

The investment opportunity set and

Received November 1990. final version received July 1992

This paper provides evidence on the cross-sectional relation between firms

Cnt~~spc~r&~~ to: Douglas J. Skinner. School of Business Administration. University of Michigan,

Elscvier Science Publishers B.V. All rights reserved

D.J. Skinner, Investment opportunities and accounting choice

empirical conclusions are as follows:

the use of income-increasing

accounting procedures is usually interpreted as

D.J. Skinner, Investment opportunities and accounting choice

2. The ios-accounting choice linkage

The ios and the accepied set qf accourzting procedures

is costly to restrict managerial choice entirely, managers choose particular

D.J. Skinner, Investment opportunities and accounting choice

6). J. Skinner. Investment oppomnities and accounting choice

More hkely to use

Firm with more

More likely to use

D. J. Skinner, hwesrment opportmifies and accounting choice

1. The Underinuestment Problem. As discussed above, while firm managers

D.J. Skinner, Investment opportunities and accounting choice

More likely to use

Firm with more

More likely IO use

Extant evidence on the debt/equity hypothesis indicates that larger

of firms I select all firms on the 1989

4. Size, leverage, and investment opportunity set proxy variables

D. J. Skinner, Incesmtent opporruniries and accounting choice

taken from annual

Compustat. 1 discuss measures

of each of the constructs

4.1. i. investment opportunity*set measures

performance because, for example, firms recent accounting performance may

4.2. Descriptive statistics

The numbers in the table are Spearman rank correlation

from zero at the 0.01 level, two-tailed

positively related to the proportion of assets that comprise assets-in-place: the

on these firms bonus plans on two levels of detail.*

Heidy (1085) and other subsequent slttdics I restrict attetiiion

to firms annual bonus

D.J. Skinner, Inaestment opportunities and accounting choice

D. J. Skinner, Investment opportunities and accounting choice

D.J. Skinner, Investment opportunities and accounting choice

The ios and accounting-based debt conuenunts

on sample firms accounting-based

D. J. Skinner, Incestntent opportunities and accounting choice

6. The relation between accounting procedure choices and size, leverage,

This section provides evidence on the determinants of three accounting

Sweeney (1991)], my results will be directly comparable to those already in the

first-in-first-out inventory cost flow assumption;

6.2. Multivariate tests

increase in the probability of choosing a more income-increasing accounting

D.J. Skinner, hestment

opportunities and accounting choice

set (10s) measures:

beta of the firms stock multiplied by the MVE/value ratio:

D. J. Skinner, hwestmenl opporrunitier and

D.J. Skinner, Inrestment

opportunities and uccounting choice

Regressions 4 and 5 in panel B of table 7 include the firm size, financial

(B) Goodwill rnerhodchoice (171 lirms)

(A) Depreciation method choice (369 firms)